Half a Million Layoffs, By the Numbers
What four years of layoff data actually shows: the peak months, the most-cut functions, the AI attribution trend, and the numbers behind the headlines.
Numbers first, narrative second. Aggregating across public trackers (layoffs.fyi, Challenger Gray reports, company filings) and our own curated event set, here is the statistical shape of the 2022–2025 correction.
The scale
Tech layoffs totaled roughly 165,000 in 2022, a peak of ~264,000 in 2023, ~152,000 in 2024, and re-accelerated in 2025 — with October 2025 alone posting the worst October for US announced cuts (all industries) in over two decades. January 2023 remains the single worst tech month: Google, Microsoft, Amazon and Salesforce announced ~48,000 cuts within twenty days of each other.
For scale: the entire correction removed roughly the headcount that US big tech added in 18 months of pandemic hiring. The industry didn't shrink below 2019 — it round-tripped the bubble, then kept trimming for different reasons.
Who got cut
Function-level analyses (LinkedIn Economic Graph, revelio-style workforce data) consistently rank recruiting/HR as proportionally hardest-hit — the function that scales with hiring died when hiring stopped. Customer support, marketing, and program/project management follow. Core engineering was cut too, but consistently below its share of headcount; security barely at all. Entry-level roles took the largest hiring-side hit: new-grad hiring down ~50% from 2019.
The AI attribution curve
In our event set, cuts with credible AI attribution went from essentially 0% of affected people in 2022, to single digits in 2023 (IBM's pause, Chegg, Stack Overflow), to roughly a fifth in 2024 (SAP, Intuit, Cisco, Dropbox, Duolingo), to approaching half of tracked people in 2025 (Amazon, Microsoft, Salesforce, Accenture, TCS, HP, Workday, CrowdStrike). Two honest caveats: attribution is partly narrative fashion (see our signal-vs-spin analysis), and the biggest AI effect — jobs never created — appears in no layoff dataset.
What the numbers don't show
Three invisible quantities matter as much as the visible ones. Suppressed hiring: teams that would have added ten now add six. Attrition absorption: the IBM model — don't fire anyone, just don't replace departures, and let AI absorb the work. And workload transfer: the same output expected from fewer people, with AI assumed to fill the gap. All three tighten the market without producing a single headline. It's why the market feels worse than the layoff numbers alone explain — because it is.
Where do you stand?
Turn the analysis into a plan — check your own exposure with the resilience calculator, or see which skills the market is rewarding.